The term digital media terminal (hereinafter “DMT”) is used herein to describe any electronic device capable of rendering media (e.g. playing back recorded or broadcasted audio/video) to a user and recording user generated media (e.g. recoding or transmitting audio/video). Digital media terminals can be found today in the markets in various embodiments, including desktop personal computers, laptop devices (such as notebook computers) and handheld devices (such as mobile “smart” phones. Such devices are becoming a preferred way of consuming media by, for example, playing downloaded or streamed audio and video content. Audio clips may range from streamed radio, to a variety of audio podcasts such as news, weather or finance reports, audio books, education, and training podcasts, music podcasts, health services and self-help podcasts, etc. In the specific case of streamed video, also known in the industry as ‘on-line video’, the content may range from short clips lasting a couple of minutes, to full length TV programs which are offered through both traditional broadcasting mechanisms and data networks as well (e.g., Internet). New ways of consuming media are becoming increasingly popular, such as “Web TV” (i.e., live streaming through Internet of TV channels), “Catch-up TV” (i.e., downloading or streaming a TV program which has been broadcasted in traditional broadcasting networks during a previous period of time), and video-on-demand (i.e. downloading or streaming any piece of video content from a catalogue).
These new media consumption trends have opened the way to new advertising markets, and therefore measuring audiences became a need in such markets.
Several methods have been attempted to track and measure this type of on-line audio/video media consumption. One solution currently used by companies that measure Web site traffic relies on software programs running at the video servers generating the video streams (or alternatively at dedicated measurement servers invoked by “tags” included in web pages) that keep track of the exact number of video sessions (also known as “site centric” approach). Such approach usually provides accurate information about media items streamed or downloaded, but provide little (if any) information about actual consumers (e.g. geographical data obtained from IP addresses).
A second approach exists by which a “software meter” is installed in a number of computers participating in a continuous survey in the form of a “panel”. Such software meters are able to detect and log internet activity, mostly in the form of WWW page visits and file downloading (hereinafter “WWW loggers”). Such approach is usually referred to as a ‘user centric’ method, since they measure the actions performed by known individuals (in terms of media consumption) as opposed to measuring activity in servers (where users may not be individualized). User centric methods resemble the methodology used in measurement of television audiences (i.e. it comprises a panel of users who accept to be monitored). WWW loggers usually capture and report many types of interactions of their users with their personal computers and the worldwide web, including streaming video and downloading of files of various kinds. Basically, such approach allows knowing who has seen which websites, for how long and in which order. User centric methods have the advantage of providing precise demographical information about users, although restricted by inherent limitations in terms of panel size and panel balance (i.e. panels must evenly represent the universes they describe, which means that all categories must be proportionally included in the panels).
Many ‘user centric’ solutions based on WWW loggers attempt to track exposure to content through the identification of URL addresses (Uniform Resource Locator) used in the World Wide Web. The URL is basically an address that specifies the location of a file on a server connected to the Internet, and it specifies the exchange protocol to be used, port number, domain name, subdirectory names, file name, etc. In order to determine what content has actually been streamed or downloaded, each URL must be univocally linked to a specific content item identified in the web space. For this reason, tables associating media items available on the web space to URLs are required before any meaningful audience figures can be obtained. These tables need to be updated every time a change is introduced, which makes the process extremely burdensome and error-prone when the number of possible items becomes too large. Moreover, such relationship between URLs and content files is not guaranteed since some advanced programming methods do not use fixed URL addresses but dynamic ones that may not bear a direct correspondence to media files, which makes such a task unfeasible.
Furthermore, new programming techniques are becoming more popular among software developers that avoid the use of URLs (beyond the initial access to a portal page), which makes the linking between URL and media files unusable.
Moreover, current implementations of WWW logger measurement services cannot track time-shifted consumption, i.e., consumption of media that has been downloaded at one point in time, and consumed at a later point in time. In such cases, most WWW loggers cannot keep track of content that has been stored and consumed at a different time; such software applications can only track Internet browsing activity and other interactions with the World Wide Web in “real time”, i.e. “as it happens”.
In order to overcome such limitations, more functionality would be required from such software meters, so that they are able to track the activity of any type of media player, with the added capability of recognizing and reporting all content consumed through the use of the digital media terminal.
To complicate things further, even streamed content—as would be the case of on-demand video services—may not necessarily imply that the associated content is been viewed during the streaming session. A large variety of software applications are available that allow recording of audio and video streaming (also known as ‘stream rippers’). Content may be downloaded (i.e. recorded) through these methods and never played back, or played back only partially. Such methods based on logging WWW activity cannot detect and report accurately this type of media consumption.
Another limitation of WWW loggers is given by the increasing amount of peer-to-peer content sharing. Once the content is downloaded from an original server, the total number of users that access the data file is impossible to track. The only viable way to track audiences to such content files is though a ‘user centric’ (i.e., panel based) approach and using a content identification technique that reports any kind of exposure to such media.
Beyond the above discussed topic of the accuracy of ‘site-centric’ and ‘user centric’ methods for the audience measurement of on-line media consumption, time shifted viewing, and peer-to-peer content sharing, an issue that is highly relevant when using WWW loggers is privacy. In order to provide an accurate report of media consumption, WWW loggers need to track all browsing activity as well as use of other software applications, sometimes including keyboard activity. This becomes a privacy issue for many users.
A theoretical solution would be available to overcome many of the problems associated to measurement systems based on activity logs through the use of content identification techniques. Such theoretical solution would require the use of complex software techniques to intercept the audio or video stream being played back by the DMT in order to attempt identification of such content through some suitable method, such as detecting an embedded audio code or watermark, or comparing signatures or fingerprints of such streams.
However, such solution is not feasible since most operating systems running on digital media terminals do not provide any standard way of accessing digital audio streams being played back by the DMT for the purpose of copyright protection. For example, Microsoft® has implemented in its latest operating system Windows Vista™ a feature called “Protected Media Path” that makes it impossible to tap into any of the audio or video digital streams produced within the system, to protect that media content from any kind of tampering. Any method based on such principles cannot be implemented in standard operating systems, or cannot be guaranteed to work on every version and all possible configurations thereof.
Even if the playback digital stream would not be protected, such a solution would suffer from other problems that would make it impractical. For example, in the Windows XP environment it is possible to write a virtual audio driver software that can intercept any audio played back through any application running in that operating system, but only if the virtual audio driver is defined as the system's “default audio device”. Such a solution cannot ensure stable and uninterrupted operation, in all possible situations and all possible configurations, because it relies on software techniques that act at the lowest layers of the system. For example, such virtual drivers may interfere with preinstalled drivers or bear incompatibilities with other installed hardware devices, or may even change audio settings of the host digital media terminal every time it is invoked, hindering its normal use as a media player or as a communications device. That, in turn, significantly increases the risk of users refusing the installation of such software in their digital media terminals, or abandoning the panels for being excessively burdened by these inconveniences.
The aforementioned limitations of all known methods seriously limit the capabilities of any user-centric approach due to the impossibility of implementing a homogenous method across all possible distribution platforms and all types of DMTs. This drawback seriously compromises the representativeness and balance of statistical samples, which is a prerequisite of any audience measurement process.